Moody's filed Q1 2026 10-Q on April 23. On the surface, a routine print: revenue +8%, adjusted operating margin +150bps to 53.2%, MA ARR +8% organic. Two items in the MIS MD&A and one in the MA Note 1 reframe what the consensus model for the NRSRO duopoly and the broader financial-data peer group should look like.

What the filing says

MIS revenue +10% to $1,140M. CFG +16%, PPIF +13%. The MD&A explicitly cites "AI-related financing from hyperscalers" (CFG) and "AI and data center-related issuance" (PPIF) as drivers. First time MCO has named hyperscaler / datacenter language as a Ratings revenue driver in a 10-Q.

MA ARR $3,607M +8% organic CC: DS +10% (KYC +13%), R&I +7%, D&I +6%. R&I had explicit "US Government attrition" language in Q1 2025; Q1 2026 has zero new attrition language and "continued strong retention." That cycle is fully lapped.

D&I segment Note 1 (the lawyer-reviewed business description) names "AI model development and risk assessment" as a customer use case — formally classifying AI labs as paying customers.

Buybacks $1.471B in Q1 alone, 73.5% of the full-year $2.0B guidance. March: 2.56M shares at $443.94 as the stock pulled back. Cash fell $915M net.

What the market thinks

MCO at 32.7x P/E, RSI 56, max pain $460, idio 23.6%. The market correctly attributes value to NRSRO + AI capex layers — MCO never participated in the AI-displacement panic that hit FDS (-46% 1Y, 14.4x P/E), MORN (-37%), RELX (-38%), CLVT (-63%).

SPGI moved -1.3% in the week after MCO's print, despite SPGI's Q4 2025 transcript (CEO Cheung, Feb 10) forecasting "significant debt issuance hyperscaler investments AI infrastructure" continuing into 2026. The cross-ticker read-through is not priced. Sell-side MIS / Ratings models treat both segments as purely credit-cycle dependent; a multi-year AI infrastructure capex layer is not in published estimates.

Why the gap exists

The hyperscaler / datacenter language is new this quarter. The "AI displacement of financial data" panic of 2024-25 compressed peer multiples (FDS now at 14x vs MCO 32x — the same business model split by sentiment). The substrate counter-thesis — NRSRO captures AI capex; data moats serve AI as input rather than get displaced — requires synthesizing lawyer-reviewed business descriptions, segment-level MD&A revenue drivers, and 6+ quarters of retention data across seven names. Cross-sector analyst coverage gaps slow it: MCO ratings analysts and FDS data analysts rarely overlap.

Risks (ranked)

1. SPGI Q1 2026 10-Q (filing in 1-2 weeks) silent on AI/datacenter language in Ratings MD&A — would force re-reading MCO's disclosure as company-specific rather than duopoly pattern.
2. Credit cycle softens — MIS transaction revenue is 67% of MIS, cyclical regardless of AI layer.
3. Q1 was issuance pull-forward not multi-year layer — pre-rate-uncertainty front-loading.
4. Any peer reports retention <90% next print — empirical refutation at corpus level.
5. Per-seat pricing compression in MA as bank headcount shrinks (managed via consumption-pricing trials).

Catalysts

• Late April – mid May 2026: SPGI Q1 earnings + 10-Q. Direct test of AI capex confirmation in own MD&A.
• ~June 2026: FDS Q3 FY26 — margin compression trajectory; gates the highest-magnitude expression of the substrate thesis.
• Q3-Q4 2026: MCO MIS CFG growth trajectory tests AI capex sustainability beyond Q1.
• 2027 10-K: MCO FY2026 buyback execution vs $2.0B guide.

What would change our mind

• SPGI Q1 10-Q omits AI/hyperscaler/datacenter in Ratings MD&A → invalidates duopoly read.
• Any financial-data peer reports retention <90% on next print → empirical refutation at corpus level.
• FDS Q3 FY26 margins worsen with management framing as structural rather than temporary.
• MIS issuance decelerates independent of credit spreads → suggests Q1 was pull-forward only.
• AI-native CRA achieves NRSRO designation (extreme tail) → regulatory moat invalidated at decade scale.

Evidence